merman-lsp

merman-lsp is the canonical LSP transport for Merman diagnostics and protocol-neutral editor intelligence. It validates the LSP projection while analysis and editor-core own the underlying language behavior.

Responsibilities

• Accept initialize, didOpen, didChange, didSave, didClose, completion, hover, completionItem/resolve, documentSymbol, definition, references, prepareRename, rename, selectionRange, foldingRange, codeAction, semanticTokens/full, and semanticTokens/range.
• Advertise editor-agnostic Merman extension requests under ServerCapabilities.experimental, including merman/ruleCatalog and merman/configSchema.
• Publish diagnostics from merman-analysis, including document pull diagnostics.
• Keep document state versioned so diagnostics from stale analysis snapshots are suppressed before publication.
• Project merman-editor-core completion, hover, document symbols, selection ranges, folding ranges, workspace symbols, definition, references, prepare-rename, rename, and semantic-token responses into LSP types.
• Provide fix-backed quickfix code actions from shared analysis diagnostics.
• Reject workspace/diagnostic while unopened workspace-file scanning has no owner; tracked document snapshots still support workspace symbols.

Deferred

• Formatting

Notes

• Plain Mermaid files and Markdown/MDX fenced Mermaid blocks are both supported.
• Diagnostics remain analysis-driven; the LSP layer does not reimplement parse or render rules.
• Language features remain editor-core-driven; the LSP layer converts URI/range/type shapes and handles request lifecycle.
• First-class family coverage is tracked in docs/lsp/CAPABILITIES.md; parse/render support outside that matrix is not yet a mature LSP commitment.